Table 1 Performance of all models: average sliding window accuracy (%) with standard deviation.
Joint-type | Classification Accuracy (%) | |||||||
|---|---|---|---|---|---|---|---|---|
LDA | Knn | FCNet | 1D-Conv | 1D-CapsNet | 2D-Conv | 2D-CapsNet | Mean | |
Joint-type accuracy | ||||||||
Left hand | 59.63 ± 0.6 | \(55.89\pm 0.5\) | \(50.15\pm 0.2\) | \(55.57\pm 0.7\) | \(55.12\pm 0.5\) | - | - | \(55.27\pm 0.5\) |
Right hand | \(51.15\pm 0.3\) | 58± 0.5 | \(51.26\pm 0.1\) | \(57.84\pm 0.2\) | \(50.32\pm 0.3\) | - | - | \(53.72\pm 0.3\) |
Hands | \(52.63\pm 0.2\) | \(54.89\pm 0.6\) | \(55.25\pm 0.3\) | \(59.19\pm 0.3\) | \(56.55\pm 0.2\) | 59.57± 0.1 | \(50.65\pm 0.1\) | \(55.53\pm 0.3\) |
Left foot | 75.63± 0.1 | 64.84 ± 0.2 | \(71.63\pm 0.6\) | \(70.1\pm 0. 3\) | \(60.89\pm 0.7\) | - | - | \(68.61\pm 0.4\) |
Right foot | \(71.31\pm 0.6\) | \(62.68\pm 0.7\) | 77.78± 0.3 | \(61.21\pm 0.1\) | \(68.24\pm 0.4\) | - | - | \(68.24\pm 0.2\) |
Feet | \(70.63\pm 0.3\) | \(63.34\pm 0.1\) | \(73.62\pm 0.4\) | 78.15 ± 0.2 | 81.15 ± 0.3 | 65.65 ± 0.3 | 86.25± 0.2 | 74.11 ± 0.3 |
Left knee | \(39.05\pm 0.3\) | 61.63± 0.2 | \(53.05\pm 0.6\) | \(58.78\pm 0.4\) | \(58.25\pm 0.1\) | - | - | \(54.15\pm 0.3\) |
Right knee | \(50.1\pm 0.2\) | 59.42± 0.1 | \(51.55\pm 0.8\) | \(59.26\pm 0.3\) | \(57.14\pm 0.1\) | - | - | \(55.49\pm 0.3\) |
Knees | \(50.55\pm 0.2\) | \(33.6\pm 0.3\) | \(51.23\pm 0.1\) | \(59.78\pm 0.1\) | 61.22± 0.7 | \(59.66\pm 0.2\) | \(60.19\pm 0.3\) | \(53.75\pm 0.3\) |
Full-body | \(39.63\pm 0.9\) | \(50.89\pm 0.5\) | \(57.88\pm 0.2\) | \(56.52\pm 0.3\) | \(60.6\pm 0.1\) | \(56.12\pm 0.3\) | 65.51± 0.1 | \(55.31\pm 0.3\) |
MEAN Classifier Accuracy | \(56.03\pm 0.4\) | \(56.52\pm 0.3\) | \(59.34\pm 0.4\) | \(61.64\pm 0.3\) | \(60.95\pm 0.3\) | \(60.25\pm 0.3\) | \(65.65\pm 0.2\) | - |
